Do home-made diets increase life expectancy in dogs?
There is a paper out on the internet that gets touted as THE evidence that feeding a home-made diet for dogs increases their life expectancy by almost 3 years.1
Webpages about alternative feeding (e.g. home-made diets in general and raw feeding, and probably other forms) signify it with titles like "THE SCIENCE MATTERS", "Diet does help dogs live longer!", "New Study: Raw Meat Diets Best for Dogs + 2003 Study" (though it seems a bit weird that a webpage dated to 2017 calls a 2003 study "new") or "Investigative study finds feeding fresh food extends the life ...".
And indeed, the paper's authors, veterinary surgeons Gerard Lippert and Bruno Sapy, have analysed the data of over 500 deceased dogs and claim to have found that dogs on a home-made diet lived, in average, about 33 months longer than dogs on an "industrial" diet, with dogs on a mixed diet in the middle.
Great, one would think. I certainly would like it, because I feed my dog raw in the good knowledge that there is almost only anecdotal evidence for the benefits and reasonably good evidence for risks.
Alas, this paper can not support the authors claim and even less so the hyperbolic presentation on a multitude of webpages.
It starts with the study design. It is an observational study and while the authors do not explicitly state it, data collection seems a mix of records from the crematorium and owner reports. An observational study can, by its very nature, not ascertain causality. Only correlation.
They could say that in their study population, longevity correlates with a home-made diet. They can not say a home-made diet caused the longer life expectancy.
Then there are methodological weaknesses. The main one is no control for confounding factors. The data also largely depends on owner reporting, which is notoriously unreliable. The authors themselves assume owners might not remember from where they got the dog, but then they assume owners correctly remember what they have been feeding the dog throughout it's life.
Extrapolating these findings to the whole dog population does not seem reasonable, as there is good evidence, as the authors themselves acknowledge, that their study population of privately cremated dogs in Belgium is not representative. Certainly using this study to advocate raw feeding is by no means justified. The "home-made diet" category in the study is defined as "with products used from the owner's meals". To me this sounds unlikely that the authors meant raw meat.
There are other, less severe issues with this "paper". There are other methodological weaknesses, data gathering and processing are not clear, the authors infer a lot of things without elaborating the why, it is actually not a paper but a "statistical essay" that's not peer reviewed.
ChatGPT
Out of curiosity, and because I am not too good at statistics, I also asked ChatGPT, first for a general evaluation of the article (to get an idea how it would see the article compared to my evaluation) and then specifically for the quality of the statistics.
The first prompt was:
"Can you evaluate the article in the following link for it's scientfic rigour and quality?
(link to paper1)"
Excerpt from ChatGPT's answer:
Scientific rigor: 4/10
Historical interest: 7/10
Reliability for making strong dietary conclusions: 3/10
(...)
Does the paper prove homemade food extends lifespan?
No.
(...)
ChatGPT also points out that newer studies into the topic with much larger animal numbers did not find this correlation.
The second prompt was:
"Can you evaluate this article for the quality of it's statistical approach?"
Excerpts from it's answer:
(...)
Data collection: 5/10
Descriptive statistics: 6/10
Inferential statistics: 4/10
Control of confounding: 2/10 (this one actually irritates me, because I can not find any control of confounding in the article)
Causal inference: 1/10
Reporting transparency: 4/10
Overall: 4/10
(...)
So, to get back to the title question, do home-made diets increase life expectancy in dogs? And are the exuberant web page titles concerning it justified?
The answer to the first question is, this paper can not answer the question. Newer research, it seems (I have not checked the studies ChatGPT quotes), does not confirm this conclusion, so the answer is probably "no".
The second question can be answered with a resounding "NO". Especially for raw feeding. This article does, and can, not justify this.
If you have any comment, feel free to drop me a line.